Polling: Unskewed Polls and comments on polling (moved from Harris Thread)

[PhillyGuy]

@DigitalC

This sounds more like just giving Trump a handicap because they were so off last time.

Giving Trump a handicap is unethical. I think it would also be highly unlikely, but I may have a higher level of social trust than the median on this board…

One thing I am seeing in these threads is greater distrust of mainstream media, and pollsters are mostly connected with that. It used to be that distrust of mainstream media was mostly a right-wing thing, but it is getting to be where both sides agree.

If both sides are correct, this would be relevant:

The most important U.S. political trend of this century is the march of college-educated White voters toward the Democratic Party . . .

And who are the employees of the pollsters paid to make decisions on how to weight by race, age, gender, education, and registration? College educated White voters!

Might someone reading this be thinking: But there is a billionaire media owner who ordered those college educated White people to ignore recent voting registration data to make Trump look competitive – and those upper middle class college educated White people just roll over and do as ordered.

If anyone thinks that, it sounds implausible to me.

[PhillyGuy]

@dtilque

I would expect a newly registered voter, one that registered in the lead-up to the election, to be more likely to vote than your typical registered voter.

Almost half the states allow same-day registration. They all vote. This is a plausible source of pollster error. I resist the idea that one can know in advance how this group will skew.

[dtilque]

@PhillyGuy

Almost half the states allow same-day registration. They all vote. This is a plausible source of pollster error. I resist the idea that one can know in advance how this group will skew.

There would have to be a lot of same-day registerers to make a difference. I couldn’t find anything recent about how many typically do this. In fact, all I could find was one ten-year-old article on it. There were only 13 states that allowed it then.

[puzzlegal]

@dtilque

I would expect a newly registered voter, one that registered in the lead-up to the election, to be more likely to vote than your typical registered voter. I’m sure they have a good reason for registering at that time.

I wouldn’t. Studies have shown that older people vote more than younger people, and it’s not because people used to vote a lot more. It seems to be that once you start voting, you are more likely to keep doing so. There’s a certain learning curve to voting.

I expect a certain number of people register, and then don’t quite manage to follow through.

@PhillyGuy

Almost half the states allow same-day registration. They all vote.

That’s different, as they register as part of voting.

[Jas09]

@puzzlegal

I expect a certain number of people register, and then don’t quite manage to follow through.

That’s certainly true, but at least in 2020 it looks like about 94% of registered voters actually vote.

In the U.S., there’s a huge gap between voting-age turnout (62.8% in 2020) and registered-voter turnout (94.1% that same year). In essence, registered voters in the U.S. are much more of a self-selected group than in other countries – already more likely to vote because, in most cases, they took the trouble to register themselves.

I can’t find any numbers about whether it is more likely to be newly-registered voters that make up the 6% that didn’t vote or previously-registered voters. My guess is it’s primarily newly-registered ones since I have seen the data you alluded to that indicate the number one indicator of likelihood to vote is previous voting record. Once you vote twice you almost always continue to vote.

Also, TIL that North Dakota doesn’t even have voter registration! They do not require proof of citizenship to vote, just proof of residency. How do they know they are all citizens??? Why isn’t the GOP flipping out about lax voter laws in North Dakota??? For some reason it’s only places like Florida and Georgia that they are up in arms about…

[LSLGuy]

As Dirty Harry didn’t quite say:

Nuthin’ wrong with votin’. As long as only the right people get to vote.

[DSeid]

@puzzlegal

I expect a certain number of people register, and then don’t quite manage to follow through.

Cite available but the data is that the vast majority of those who register in an election cycle do follow through and vote in it. I would not be shocked if it is a higher percent than those who tell pollsters that they will vote who follow through.

[puzzlegal]

@DSeid

Cite available but the data is that the vast majority of those who register in an election cycle do follow through and vote

Well, that’s wonderful news.

@Jas09

Also, TIL that North Dakota doesn’t even have voter registration! They do not require proof of citizenship to vote, just proof of residency.

My liberal state has voter registration, but does not require proof of citizenship. You need to legally swear or affirm your citizenship, and you are reminded that lying about that is a serious crime. I have not heard any credible stories of non-citizens registering.

[Red_Wiggler]

Anecdata from an election judge who directly processes same day registrations: those voters make up fewer than 1% of all voters in the precincts I’ve worked. I live in a red suburban county that probably skews a little older than the national average, ymmv.

[Win_Place_Show]

I think it can be said (“cite unseen”) that most mail-in ballots skew Democrat, correct?

I wonder if there are any studies done to identify whether early voters tend to skew either Democrat or Republican. Then we could bump that up against early voting numbers thus far (as in “after one week of early voting, xx,xxx voters have gone to the polls in Pennsylvania, compared to yy,yyy voters who did during the first week of early voting in 2020”).

Granted 2020 might not be the best “apples to apples” comparison, but it would be interesting to crunch the numbers thus far to try and glean if the enthusiasm (or lack thereof) in Pennsylvania so far is a good sign or a bad sign.

[DrDeth]

@dtilque

I would expect a newly registered voter, one that registered in the lead-up to the election, to be more likely to vote than your typical registered voter.

I certainly agree. But generally, iirc Pollsters work from a list of registered voters, and that list may not be up to date?

[Leaper]

So do downballot races, including initiatives, have any relevance to the presidential race numbers? We often look at the relationship both ways, but I perhaps erroneously detected less discussion of that direction this time around.

[DSeid]

@CaveMike

Meaning there was a 3% error due to systematic issues (like over- or under-represented sub-populations) that is independent of the margin of error from sampling?

I’m not actually I quite understand the question but suspect this bit from 538 answers it. There are also some bits that are worth quoting to highlight the limitations of the tool.

How to read political polls in 2024

Since 1998, polls of presidential general elections have had a weighted average error of 4.3 points; of Senate general elections, 5.4 points; of House general elections, 6.1 points. When a candidate leads in the polls by a smaller margin than that, be aware that it would be totally normal by historical standards for the trailing candidate to win.
… snip …
although polls of a given election do often end up being too good for one party or the other (see “polls are generally accurate — but not perfect” above), it’s virtually impossible to guess in advance which direction that bias will run in. Historically, polling bias has jumped around at random from cycle to cycle.

So actually +/- 4 systemic bias in a presidential cycle is typical and it swings both ways. I don’t know about “random” but not past results predicting performance future results.

[ShadowFacts]

This is one of the most encouraging articles I’ve read about the election in a while, at least in the electoral sense. NY Times gift link: " Republicans’ Electoral College Edge, Once Seen as Ironclad, Looks to Be Fading"

There’s a decent amount of hedging in the article, but the upshot is:

But there’s growing evidence to support a surprising possibility: His once formidable advantage in the Electoral College is not as ironclad as many presumed. Instead, it might be shrinking.

That EC advantage is the thing that most worries me about the upcoming election. Even as polling and vibes show Harris slightly ahead in many places, "slightly "is not nearly enough if she needs to be ~4 pts ahead to actually win the EC. If that disadvantage in significantly ameliorated as this article suggests, then I feel a lot better about her chances to actually make it to the White House.

[JKellyMap]

Yes. If this is true, it’s likely because she’s polling worse than Biden ‘20 or Clinton ‘16 among Blacks and Latinos (especially, young Black men), but polling better than expected among many Whites, especially women….

….and White women are geographically distributed in a way that reduces Harris’ EC disadvantage. (Namely, there’s a ton of them in PA, WI, and MI).

[DSeid]

In the current aggregated trend lines PA, the likeliest tipping point state, is Harris +1.3 and national is Harris +2. EC advantage -0.7.

Recent same dates Morning Consult for complete apples apples PA H+2 and national H+4. EC advantage -2.

Error bars apply but systemic error potential and direction same for both, whatever it is.

Of course state polls can be off differently than national ones. But that is how it is running.

And @JKellyMap because she is fighting for rural voters hard, especially in PA, possibly decreasing his margin there.

[puzzlegal]

@JKellyMap

.and White women are geographically distributed in a way that reduces Harris’ EC disadvantage. (Namely, there’s a ton of them in PA, WI, and MI).

I have a friend who thinks the “don’t want to tell pollsters she is pro abortion rights” error is going to be larger than the “don’t want to tell pollsters he is pro trump” error was in trump v clinton.

[veryfrank]

@Win_Place_Show

I think it can be said (“cite unseen”) that most mail-in ballots skew Democrat, correct?

The Republicans used to like mail-in voting because it’s how the members of the military posted overseas would get their ballots in, and there’s a tendency (how big I don’t know off the top of my head) for military people to lean conservative. They only turned against it during the 2020 election as more people needed/wanted to mail-in because of the pandemic, and the Trump Party objected to the levelling of the playing field. Plus TFG got it into his head that it was rife with fraud and the party followed suit.

[digs]

I’m a little worried about the recent CNN poll showing a dead heat.

Can someone talk me off the ledge?

[Jas09]

@digs

Can someone talk me off the ledge?

Not really. Quinnipiac has the same result. And they ended 2020 with Biden+11. PA also has 6 recent polls that have a tie or Trump barely ahead. That is offset by a handful of polls showing a modest Harris lead.

It is tight, and Trump is probably a bit ahead in the EC.

Harris needs to finish strong and have the better ground game on election day. There is a reason she wants another debate - she needs a way to keep positive momentum because the “steady-state” of the Trump-Harris matchup appears to be a national tie or small Harris lead and a small Trump lead in enough states to make it a virtual tie in the EC.